VOGONS


First post, by byte_76

User metadata
Rank Member
Rank
Member

I have a dead GTX 1070 that had two blown mosfets.

The first one was an AOE6930 which had no signs of physical damage but was shorting the 12V rail so I replaced it.
The second was a V3605 located on the end of the card and was clearly burnt.

I'm not very knowledgeable on the electronic workings of a graphics card, so I just looked for damage and shorts and fixed what I found. (Initially I just removed the two mosfets and then confirmed that the PC starts with the card installed as secondary and that the card is at least detected but otherwise non-functional)

When I powered the card after replacing the mosfets, without connecting the auxiliary power (ie. only plugged into the motherboard), the V3605 burned out again.

I need help to identify the cause of the V3605 burning out. (What is this mosfet's purpose on the board?)

I have circled the component that is burnt in the image below.

Please also assist me to identify the measuring points for Core, Memory, PEX, 1.8V, 3.3V and 5V with the approximate readings I should be expecting.

Reply 1 of 7, by paradigital

User metadata
Rank Oldbie
Rank
Oldbie

Something is causing that MOSFET to blow by drawing too much current through it.

When you removed it, was the short gone?

Reply 2 of 7, by byte_76

User metadata
Rank Member
Rank
Member

The 12V was shorted by a single mosfet located on the left side of the image attached below. (indicated with a green arrow)
Removing that mosfet did remove the short but when I replaced the mosfet and powered the card with the auxiliary power, the new mosfet failed again and resulted in the 12V being shorted again.

I have not found which rail might be shorted by the small mosfet located on the right side of the image (Near the edge of the card. Indicted with the green arrow).
That mosfet burnt out while the card was installed into the slot but not connected to the AUX power.
I'm not sure if the rail that it belongs to is shorted because I first need to identify which rail it belongs to and where to measure that rail for shorts.

Last edited by byte_76 on 2025-09-30, 17:29. Edited 2 times in total.

Reply 3 of 7, by byte_76

User metadata
Rank Member
Rank
Member

I've once again removed the two mosfets indicated by the green arrows in the post above and now the PC turns on with the card installed and the AUX power connected.
The core gets hot over time (about 1 minute) but there is no display output.

If I set the BIOS to use the onboard display and boot into Linux, the GTX 1070 is detected but MATS indicates that all memory phases are in an error state.

I only have 4 more of each mosfet available and I'm quite sure if replace them a second time, they will fail again but I don't know why.

I have not removed the mosfet driver for the mosfet on the left of the image. I'm not sure if it should be replaced together with the mosfet and I don't have any spares available.
Do you think the card should work with the mosftets removed, albeit slightly crippled in terms of power.

Reply 4 of 7, by Unknown_K

User metadata
Rank Oldbie
Rank
Oldbie

If that Mosfet is for the RAM then the card will not work.

Collector of old computers, hardware, and software

Reply 5 of 7, by byte_76

User metadata
Rank Member
Rank
Member

I've tested all the voltages, I've got 12V, 5V, 1.8V and PEX.

My DM just indicates ohms when I check the memory inductor while the card is powered. Without power, the reading is 108ohms.
What voltage should I be reading for memory while the card is powered?

I don't have a thermal camera but it might have been helpful to identify a shorted module.

I assume there are enough redundant mosfets that the card should still work with those that I've removed.

Reply 7 of 7, by byte_76

User metadata
Rank Member
Rank
Member
tehsiggi wrote on 2025-10-05, 16:59:

Looks like the reference PCB of the 1070, so perhaps this 1070Ti (which had the same PCB iirc) schematic could help..

Thank you!